摘要
光谱和空间信息的联合使用是高光谱图像分类领域的研究热点之一.本文在已有的矩阵判别分析(MDA)模型的基础上,提出了一种基于稀疏图正则的改进模型.在有效融合高光谱图像光谱-空间信息的同时,能充分挖掘无标签样本的信息,从而提升了模型的分类性能.为了验证本文算法的有效性,在两个高光谱数据集上,与多种方法进行了对比.实验结果表明,本文提出的算法优于其他同类算法.
In the field of hyperspectral image classification,the incorporation of spectral information and spatial information is one of the hot research topics. In this paper,a modified matrix-based discriminant analysis(MDA)model is proposed based on sparse graph regularization. The proposed model can not only combine the spectral information and spatial information effectively,but also sufficiently explore the wealth information from unlabeled samples,thus improving the classification performance. In order to verify the effectiveness of the proposed method,experiments have been conducted on two widely used hyperspectral images. The experimental results show that the performance of our method is superior as compared to other methods.
作者
黄晓伟
杭仁龙
孙玉宝
刘青山
Huang Xiaowei;Hang Renlong;Sun Yubao;Liu Qingshan(Jiangsu Key Laboratory of Big Data Analysis Technology,School of Information and Control,Nanjing University of Information Science and Technology,Nanjing 210044,China)
出处
《南京师大学报（自然科学版）》
CAS
CSCD
北大核心
2019年第1期51-58,共8页
Journal of Nanjing Normal University(Natural Science Edition)
基金
国家自然科学基金(61672292)
江苏省高校自然科学研究面上项目(18KJB520032)
江苏省青年基金项目(BK20180786)
关键词
高光谱图像分类
谱-空特征融合
矩阵判别分析
稀疏图正则
hyperspectral image classification
spatial-spectral feature fusion
matrix-based discriminant analysis(MDA)
sparse graph regularization
